Annu. Rev. Astron. Astrophys. 1998. 36:
435-506 Copyright © 1998 by Annual Reviews. All rights reserved |
Photometric and spectroscopic techniques can be used to estimate heavy-element abundances in Local Group dwarfs. For early-type galaxies, the properties of the RGB constrain [Fe/H] and, in some cases, the abundance dispersion, [Fe/H]. Apart from helium and some molecular species, photometry is poorly suited to determining abundances of individual elements. Photometric abundances have now been measured in some Local Group dIrr systems where the old/intermediate-age RGB population can be observed directly in deep color-magnitude diagrams (e.g. Sextans A: Dohm-Palmer et al 1997; Leo A: Tolstoy et al 1998; various: Lee et al 1993c; see Table 6 for other examples). Spectroscopy can be used to determine abundances of both individual stars and emission nebulae, such as HII regions and planetary nebulae. Generally, spectroscopy provides abundances for specific elemental, ionic, or molecular species, which somewhat complicates comparisons with photometric [Fe/H] abundance estimates. Some good recent reviews of dwarf abundances can be found by Skillman & Bender (1995), Skillman (1998).
5.1. The Observational Basis for Local Group Dwarf Abundances
RED GIANT BRANCH ABUNDANCES Da Costa & Armandroff (1990) observed RGB sequences using the V and Cousins I bands in a number of globular clusters ranging from -0.7 to -2.3 in [Fe/H]. These sequences have helped establish the (V-I) TRGB method of determining distances (Section 2.2). They also found that the colors of the giant branches define a monotonic sequence with respect to abundance: [Fe/H] = -15.16 + 17.0( V - I)-3 - 4.9(V - I)-32, where (V-I)-3 is the reddening-corrected (V-I) color of the giant branch at MI = - 3.0. This relation is valid for - 0.7 > [Fe/H] > - 2.2. Lee et al (1993c) reevaluated this expression at MI = - 3.5 for easier application in distant galaxies: [Fe/H] = - 12.64 + 12.6(V-I)-3.5 -3.3(V-I)-3.52. This relation has the same range of validity as the earlier one because both were derived from the same data (Da Costa & Armandroff 1990).
Photometric abundance indicators such as the RGB color run into complications in dwarf galaxies. Unlike clusters, most dwarfs exhibit significant abundance dispersions. Because dwarf galaxies are considerably more distant on average than Galactic globular clusters, few galaxies have reliable HB photometry (though see Da Costa et al 1996); those that do often reveal unusual HB morphologies, such as bimodal (in luminosity) HB sequences (Carina: Smecker-Hane et al 1994; Sagittarius: Sarajedini & Layden 1995) or compact, super-red HBs (Leo I: Lee et al 1993a, Caputo et al 1995). It is not always feasible to use abundance indices such as (B-V)0,g (the B-V color of the RGB at the level of the HB; Sandage & Wallerstein 1960, Sandage & Smith 1966, Zinn & West 1984) to estimate the metallicities of these galaxies. Another complication is that most Local Group dwarfs exhibit complex star-formation histories (Section 6). It remains to be seen how reliably these abundance indices - derived for globular clusters - can measure the composite populations of nearby dwarfs (see also Grillmair et al 1996). It would clearly be useful to extend the range of photometric abundance indicators with observations of the RGBs in populous clusters in the Galaxy and the Magellanic Clouds.
With few exceptions, every Local Group dwarf for which abundances have been determined from the RGB shows evidence of a significant abundance dispersion (Table 6). In M32 (Grillmair et al 1996) and NGC 205 (Mould et al 1984), the color distribution of the RGB implies an abundance distribution that is skewed toward higher abundances. Although these color dispersions principally reflect variations in abundances, an age dispersion can also (slightly) broaden the RGB (e.g. Bertelli et al 1994, Meynet et al 1993). Carina has a large age spread and populations with distinct abundances (Da Costa 1994a, Smecker-Hane et al 1994), yet it has a narrow RGB. Leo I also exhibits a large spread in age but has a wide giant branch (Lee et al 1993a). In Carina, the age spread seems to compensate for the metallicity dispersion, while in Leo I, it does not. Does this imply radically different chemical-enrichment patterns for the two galaxies? This underscores an obvious, but important, point: The star-formation and chemical-enrichment histories of dwarfs cannot be interpreted independently. To derive one history requires careful consideration of the other (Hodge 1989, Aparicio et al 1997b, c).
SPECTROSCOPIC ABUNDANCES Among the early-type dwarfs in the Local Group, spectroscopic abundances (typically [Fe/H] or [Ca/H]) have been measured for individual stars in Draco (Lehnert et al 1992 and references therein), Sextans (Da Costa et al 1991, Suntzeff et al 1993), Carina (Da Costa 1994a), Sagittarius (Da Costa & Armandroff 1995, Ibata et al 1997), and Ursa Minor (EW Olszewski & NB Suntzeff, private communication). Oxygen abundances have been derived from spectroscopy of planetary nebulae in Fornax (Maran et al 1984, Richer & McCall 1995), Sagittarius (Walsh et al 1997), NGC 185, and NGC 205 (Richer & McCall 1995), as well as the dIrr NGC 6822 (Dufour & Talent 1980).
Spectroscopy of HII regions in Local Group dIrr galaxies typically targets oxygen, but abundances of other elements such as nitrogen, sulphur, and helium have also been measured (e.g. Pagel et al 1980;, Garnett 1989, 1990;, Garnett et al 1991). The improved blue-sensitivity of CCD detectors in recent years has greatly aided nebular abundance studies by making it simpler to derive reliable physical conditions in HII regions. This alone greatly improves the consistency and precision of the oxygen abundances derived in this manner (see Skillman 1998 for details). Table 6 lists the oxygen abundances for Local Group dIrrs that have adequate data. Unlike [Fe/H] in early-type dwarfs, there is no evidence for significant dispersion of the oxygen abundances in any Local Group dIrr in which multiple HII regions have been studied (Pagel et al 1980, Skillman et al 1989a, Moles et al 1990, Hodge & Miller 1995). Because HII regions are associated with young populations (though the gas itself may derive from relatively old stars), these nebular abundances nicely complement those derived from intermediate-age planetary nebula (Olszewski et al 1996b).
Integrated spectroscopy is impractical for most of the dwarfs largely because of their extremely low surface brightnesses [Sembach & Tonry (1996) suggested one method to overcome this problem]. M32 is a famous exception for which a number of integrated spectroscopic studies in the optical and ultraviolet (UV) have been carried out (for reviews, see O'Connell 1992, Grillmair et al 1996). The excess UV light in its spectrum is generally taken as evidence of a population of relatively young stars associated with the IR-luminous asymptotic giant branch (AGB) stars found by Freedman (1992), Elston & Silva (1992). However, the spectrum only weakly constrains the stellar abundance of M32 (Grillmair et al 1996). Optical spectra reveal complex radial gradients of the Balmer lines (these weaken outward), while the Mg lines remain constant with radius and CH increases in strength (Davidge 1991, González 1993). Oddly, M32 exhibits no strong radial color gradients apart from in the UV (Michard & Nieto 1991, O'Connell 1992, Peletier 1993, Silva & Elston 1994). These conflicting tendencies have greatly complicated spectroscopic determinations of the galaxy's abundance even now that deep HST photometry is available (Grillmair et al 1996). The nuclear region of NGC 205 has also been observed spectroscopically (Bica et al 1990), revealing a prominent young population (age 108 years) with a mean metallicity of [Fe/H] ~ -0.5, along with more older, metal-poor stars (ages 5 Gyr; [Fe/H] -1.0). A lower nuclear abundance has been derived by Jones et al (1996) for NGC 205 ([Fe/H] ~ -1.4) from UV spectra.
5.2. The Metallicity-Luminosity Relation and the dIrr/dSph Connection
The fact that the more luminous dwarf galaxies are also on average the most metal rich has been known for some time for both dIrr (Lequeux et al 1979, Talent 1980, Skillman et al 1989a) and dSph galaxies (Aaronson 1986, Caldwell et al 1992). Aaronson (1986), Skillman et al (1989a) merged the abundance data for both types into a a single luminosity-abundance (L-Z) relation spanning 12 mag in MB and about 1.6 dex in oxygen/iron abundance. A recent determination of [Fe/H] of a low-surface brightness but relatively luminous dSph galaxy in the M81 group (Caldwell et al 1998) demonstrates clearly that luminosity, not surface brightness, is the principal parameter correlated with metallicity in dSph and, presumably, dIrr galaxies.
Figure 7 is a plot of the mean abundances for all of the galaxies in Table 6 (except DDO 210 for which [Fe/H] is uncertain; Greggio et al 1993) vs their mean V-band absolute magnitudes (Table 4). The data have been corrected for external reddening; internal extinction is probably insignificant for most of these systems (see Table 5). In Table 6, 10 galaxies have reliable [Fe/H] and oxygen abundances, including some dIrrs with stellar abundance estimates and some early-type dwarfs with oxygen abundances from planetary nebulae. The mean difference between [Fe/H] and [O/H] [defined as log(O/H) - log(O/H)] is 0.37 ± 0.06 dex. I have therefore added -0.37 to the oxygen abundances before plotting them in Figure 7.
Figure 7. A plot of [Fe/H] (filled squares) or [O/H] - 0.37 [I assume 12 + log(O/H) after Anders & Grevesse (1989)] vs absolute V-band magnitude. The dotted line is a rough fit to the [Fe/H]-MV relation for the dSph and transition objects. Sagittarius corresponds to the points near (MV,[Fe/H]) ~ (-13.4,-1.0). Square symbols refer to dSph or dE galaxies; triangles refer to transition galaxies (denoted dIrr/dSph in Table 1); circles refer to dIrr systems. Filled symbols correspond to [Fe/H] abundances determined from stars, while open symbols denote oxygen abundance estimates from analyses of HII regions and planetary nebulae. See Table 6 for details. |
Even if the offset to the nebular abundances is disregarded, the stellar [Fe/H] abundances show a bimodal, or possibly discontinuous, behavior. The oxygen abundances only reinforce this conclusion. The "upper branch" in Figure 7 (denoted by the dotted line) contains only dSph galaxies, and all four transition galaxies (LGS 3, Phoenix, Antlia, and Pegasus; see Table 1) with precise abundance estimates. All of the galaxies fainter than MV ~ - 13.5 with nebular abundances are dIrr systems and clearly fall below the dSph relation by about 0.6-0.7 mag. At the other extreme, nearly all of the galaxies brighter than MV ~ - 13.5 are dIrr systems - with the important exceptions of NGC 147, NGC 185, NGC 205, and M32. A similar bifurcation of the luminosity-abundance (L-Z) relation was suggested by Binggeli (1994), Walsh et al (1997). Caldwell et al (1992, 1998) adopted a single relation to fit all the data for early-type dSph and dE galaxies in the Local Group and other groups and clusters. However, a single linear relation in Figure 7 ignores the strong segregation of dIrr and dSph for MV -13.5. The proposed bimodal L-Z relation also helps remove the abundance anomaly exhibited by Sagittarius (Ibata et al 1994, Mateo et al 1995c, Sarajedini & Layden 1995, Ibata et al 1997) for which the L-Z relation of Caldwell et al (1992, 1998) implies that MV, Sgr -16. This is about 2-3 mag brighter than observed (Mateo et al 1995c, Ibata et al 1997). Based on the current data, there is no correlation between the offset (in magnitudes) from the dashed line in the upper panel of Figure 7 and the intrinsic color of the galaxy (Skillman et al 1997), as might be expected if the bimodal behavior simply reflects the effects of current star-formation on the integrated luminosities of dIrr galaxies.
Figure 7 addresses the relationship between dwarf ellipsoidal (dSph and dE) galaxies and dIrr systems. Star-formation (Dekel & Silk 1986, Babul & Rees 1992, De Young & Heckman 1994) and ram-pressure stripping (Faber & Lin 1983, van den Bergh 1994c) have been proposed as means of removing gas from dwarfs in the inner Galactic halo. The spatial segregation of dSph and dIrr (e.g. see Figure 3) appears consistent with the latter idea. But there are other fundamental problems if dSph galaxies are supposed to be simply "gas-free" dIrr (see reviews by Binggeli 1994, Ferguson & Binggeli 1994, Skillman & Bender 1995). Hunter & Gallagher (1985), Bothun et al (1986) showed that the present-day central surface brightnesses of dIrrs will be considerably lower than in dSph galaxies after evolutionary fading, while James (1991) found large systematic structural differences between Virgo dIrr and dSph galaxies that seem inconsistent with a common origin or a single evolutionary endpoint (see also Section 3.2). Binggeli (1994), Richer & McCall (1995), Walsh et al (1997) all noted that at a given luminosity, dIrr galaxies are generally more metal poor than dSph systems, which is precisely the effect seen in Figure 7.
If dIrr and dSph galaxies do indeed represent fundamentally different objects, why does a protodwarf galaxy choose one type rather than the other? It is interesting that even the lowest-luminosity dIrr shows evidence for rotation (e.g. GR 8: Carignan et al 1990; Leo A: Young & Lo 1996) even though vrot / <1.0. The only rotating dSph systems are NGC 147 (Bender et al 1991) and UMi (Hargreaves et al 1994b, Armandroff et al 1995). The latter's rotation may reflect streaming motions induced by external tides (Oh et al 1995, Piatek & Pryor 1995), and both galaxies have vrot / 0 < 1.0. Could angular momentum be the factor that distinguishes dIrr (high angular momentum) and spheroids (low angular momentum)? Alternatively, Skillman & Bender (1995) suggested that the strength of the first star-formation episodes dictates this distinction; galaxies experiencing little or no early star formation become dIrr systems (see also Aparicio et al 1997c). Or is environment the deciding factor after all (van den Bergh 1994c; Figure 3)?
There are serious objections to each possibility. Three of the four early-type dwarf satellites of M31 that lie in the proposed dIrr branch in Figure 7 (NGC 147, NGC 185, and NGC 205) do not significantly rotate (Bender & Nieto 1990, Held et al 1990, 1992, Bender et al 1991), although dIrr galaxies of similar luminosity do (see Section 7). The fourth galaxy, M32, does rotate (Tonry 1984, Dressler & Richstone 1988, Carter & Jenkins 1993), but its structural parameters are not like any dIrr. Many dIrr galaxies do have pronounced ancient populations, while some dSph galaxies exhibit evidence for few or no old stars (Section 6). Thus, the amplitude or timing of the first episodes of star formation do not appear to differentiate dIrr and dSph galaxies. Finally, although galaxy types are segregated as a function of distance from M31 or the Milky Way (Figure 3), there are glaring exceptions. The Magellanic Clouds are dIrr galaxies that lie close to the Milky Way, while Tucana is an example of an inactive dSph far from any large galaxy. Environment also offers no easy explanation of the segregation of faint dSph and dIrr galaxies in Figure 7.